Orthogonal polynomials for the weakly equilibrium Cantor sets

arXiv:1506.05734vl [math.SP] 18Jun2015

ORTHOGONAL POLYNOMIALS FOR THE WEAKLY
EQUILIBRIUM CANTOR SETS

GOKALP ALPAN AND ALEXANDER GONCHAROV

Abstract. Let A’(7) be the weakly equilibrium Cantor type set introduced
in m- It is proven that the monic orthogonal polynomials Q 2 S with respect
to the equilibrium measure of K{'y) coincide with the Chebyshev polynomials
of the set. Procedures are suggested to find Qn of all degrees and the cor¬
responding Jacobi parameters. It is shown that the sequence of the Widom
factors is bounded below.

1 . Introduction

Cantor sets appear as supports of spectral measures for the Jacobi operators in
some natural situations. For example, the Schrodinger operator with an almost
periodic potential (for example, the almost Mathieu operator) typically has such
spectrum (see e.g. the review [H])- On the other hand, we can meet Julia sets K
of the Cantor type in the theory of orthogonal polynomials (OP) with respect to
the equilibrium measure fiK (see e.g [SElIl]). Whereas, due to B. Simon et ah,
there is a comprehensive theory for the finite gap Jacobi matrices, until now there
is no such theory for purely singular continuous measures. In order to analyze the
infinite gap Jacobi matrices, the class of Parreau-Widom sets was suggested (see e.g.
0 ). This notion includes Cantor sets of positive Lebesgue measure, so the Szego
condition can be applied in this case. Recently numerous attempts were undertaken
to investigate the spectral problem for self-similar measures with Cantor supports.
By means of computational methods, some conjectures (see e.g. [TT1[T2]) about the
asymptotic behavior of the Jacobi parameters and other related values were posed.

Here we consider one more example of a family of OP with respect to fiK{-y) on
the Cantor set K^j), introduced in [10]. The set K{j) depends on a parameter
7 = (7s)^i and is constructed by means of a recurrent procedure. If 7 ^ are
not very small then the set ^( 7 ) is not polar. At least in known cases, the set
K{'y) is dimensional and, by [T], fJ-K(j) is mutually absolutely continuous with
the corresponding Cantor-Hausdorff measure. This is not valid for geometrically
symmetric zero Lebesgue measure Cantor sets, where, by [13] and followers, these
measures are mutually singular.

In Section 2 we show that the monic OP Q 2 ^ coincide with the corresponding
Chebyshev polynomials. In Sections 3 and 4 we suggest a procedure to find Qn for
n ^ 2^*. This allows to analyze the asymptotics of the Jacobi parameters (an))))Li'
Since Cap{K{j)) is known, we estimate (Section 5) the Widom factors Wn ■=
check the Widom condition, that characterizes the Szego class of
Jacobi matrices in the finite gap case. In the last section we discuss a possible
version of the Szego condition for singular continuous measures. At least for 7 ^ <

2010 Mathematics Subject Classification. Primary 42C05, 47B36; Secondary 31A15.

1

2

GOKALP ALPAN AND ALEXANDER GONCHAROV

1/6, s G N, the Lebesgue measure of the set K{j) is zero, so it is not a Parreau-
Widom set.

For the basic concepts of the theory of logarithmic potential see e.g log
denotes the natural logarithm, Cap{-) stands for the logarithmic capacity, 0° := 1.

2. Orthogonal Polynomials

Given a sequence 7 = ( 7 s)//,i with 0 < 7 s < 1/4 define tq = 1 and Vs = 7sr's-i-
Let

( 2 . 1 ) Pi := X - 1 and P 2 «+i(x) := P 2 s(x) • (P 2 «(x) + Ts)

for s G No in a recursive fashion. Thus, P 2 {x) = x ■ {x — 1) for each 7 , whereas,
for s > 2, the polynomial P 2 a essentially depends on the parameter 7 . For s G Nq
consider

Es = {x G M : P 2 s+i(x) < 0} = P 2 ‘> + 1^ ([~1) 1]) ~ Gj_i/j^s,

where Nq := N U {0} and are closed basic intervals of the s—th level which are
necessarily disjoint. Then Pg+i C Eg and the set if ( 7 ) := PI/TqPs is a Cantor set
by Lemma 2 in m-

Let stand for the length of /ys when we enumerate them from the left to the
right. By Lemma 6 in |10) .

71 .. . 7 s < k,s < exp ^ 16 ^ 7 fe^ 7 i .. . 7 s, 1 < * < 2 ®,

provided 7 ^ < 1/32 for all k. Therefore, the Lebesgue measure \K{'-f)\ of the set
with this condition is zero and, by [S], if ( 7 ) is not a Parreau-Widom set. For
the definition of Parreau-Widom sets see e.g. [5]. In Section 4 we will show that
|if( 7 )| = 0 as well if 7 fe < 1/6 for all k.

On the other hand, by choosing ( 7 fc)/Ti sufficiently close to 1/4, we can obtain
Cantor sets with positive Lebesgue measure. What is more, in the limit case, when
all 7 fe = 1/4, we get Eg = [0,1] for all s and if ( 7 ) = [0,1] (see Example 1 in [TU]L
In addition, by Corollary 1 in [10], Cap{K{'^)) = exp (^^1 2 “^log 7 fe). In the
paper we assume C'ap(if ( 7 )) > 0. Let PKi-y) denote the equilibrium measure on the
set, II • II be the norm in the corresponding Hilbert space. From Corollary 3.2 in [1]
we have pK('y){Ij,s) = 2“® for all s and 1 < / < 2®, provided 7 fc < 1/32 for all k.

From now on, by Qn we denote the monic orthogonal polynomial of degree
n G N with respect to The main result of this section is that, for n = 2® with

s G No, the polynomial Qn coincides with the corresponding Chebyshev polynomial
for if ( 7 ). The next two theorems will play a crucial role.

Theorem 2.1 (|10j. Prop.l). For each s G Nq the polynomial P 2 « -I- rs/2 is the
Chebyshev polynomial for if ( 7 ).

Remark 2.2. Only the values s G N were considered in [10]. But, clearly, for s = 0
the polynomial Pi-|-l/2 = x — l/2is Chebyshev.

Remark 2.3. Since real polynomials are considered here and the alternating set for
P 2 '> + rs/2 consists of 2® -|- 1 points, the Chebyshev property of this polynomial
follows by the alternance argument.

ORTHOGONAL POLYNOMIALS FOR THE WEAKLY EQUILIBRIUM CANTOR SETS 3

Theorem 2.4 f[16). T.3.6). Let if C K 6e a non-polar compact set. Then the
normalized counting measures on the zeros of the Chebyshev polynomials converge
to the equilibrium measure of K in the weak-star topology.

For s S N, the polynomial + Ts / 2 has simple real zeros which are

symmetric about x = 1/2. Let us denote by Vg the normalized counting measure
at these points, that is Vg = 2“^* X]fc=i

Lemma 2.5. Let s > m with s,m G No- Then f (P 2 '" + di^g = 0.

Proof. For m = 0 we have the result by symmetry. Suppose m > 1. By (EU , at
the points ixk)1^i we have

-Pjs + — = (P2»-i)^ + ^s-l-f2s-l ~2 ~

The discriminant of the equation is positive. Therefore, the roots satisfy

(^23-1 + ag_]^)(P2'»-i + cis-i) — Oj

where al_i + a^_i = Vg-i and 0 < al_i,ag_i < r^-i. Thus, a half of the points
satisfy P 2 »-i + Cs-i = 0 while the other half satisfy P 2 '*-i + Cs-i = 0.

Rewriting the equation P 2 =-i + Q^s-i = 0, we see that

P2a-2 + r g—2P2‘‘-'2 + Oig—i — 0 -
Since rl _2 > 4rs_i > 4ag_i, this yields

(P2'»-2 + CIs-2)(-P2'>-2 + otl_2) — 0

with al _2 + Q;s _2 = ^s -2 and 0 < as_ 2 ,Q!s -2 < ^s- 2 - By the same argument, the
second half of the roots satisfy

( P 2'»-2 + Q ; s _ 2 )( P 2*-2 + ^ 8 - 2 ) = 0

with al _2 + as -2 = ’'s -2 and 0 < a^g_ 2 ,at -2 < Ts- 2 -

Since at each step r‘j_i > An we can continue this procedure until obtaining
P 2 m+i. So we can decompose the Chebyshev nodes {xk)l^i into groups.

All 2 ™+^ nodes from the i—th group satisfy

P2m+l + — 0, 0 < (Xm+1 < Tm+1.

By using these 2®“™“^ equations we finally obtain

(P 2 ™ + Urn ^)(P2"* + (Tm) — ^

where a‘^~^ = ^m- Thus, given fixed i—ih. group, for 2™ points from the group

we have P 2 ™ = — whereas for another half, P 2 m = —aff. Consequently, we

have

P 2

^dvg + lift =
2

2'"(-c

□

Lemma 2.6. Let 0 < ii < i 2 < .. .in < s. Then

4

GOKALP ALPAN AND ALEXANDER GONCHAROV

(a)

dl^s —

(b)

(-irllT-

k^l

■ ■ {P2*n + dUs = 0.

P2ir^ dVs

Proof. (a) Suppose that ii > 1. As above, we can decompose the nodes ixk)1‘'^i
into equal groups such that the nodes from the j—th group satisfy

an equation

(P2n +ar')(P2n +<') = 0

with If, on some set, {P 2 k +a){P 2 k +/3) = 0 with a + [3 =

Tk, then P 2 k+i = P^k +P 2 k Xk = —aj3 and each of the next polynomials P 2 '=+*
is constant on this set. Therefore the function P 2»2 ■ ■ • ^ 2 *" takes the same
value for all Xk from the j—th group. This allows to apply the argument
of Lemma 1^31

J P2'lP2'2 ■ ■ ■ P2*ridVs — -^ J P2'2P2'3 ■ ■ ■ P2'ndr's.

This equality is valid also for = 0 since

J ^Pl + 2^ -^2*2 ■ • ■ P2'r<. dVs = 0,

by symmetry. Proceeding this way, the result follows, since —Traj^ =
f P 2 mdi's, by Lemma [2.51

(b) Opening the parentheses yields

/ Tl n

P2n P2-2 . . . P2^. + E T / n ^2-.- '

k=l j^k

By Lemma l2.5l and part (a), this is

k—1 k—0 ^ ^

fc=i

□

Remark 2.7. We can use HK(-y) instead of z/s in Lemma [2.51 and Lemma [2.61 since,
by Theorem 12.41 in the weak-star topology.

Theorem 2.8. The monic orthogonal polynomial Q 2 ‘ with respect to the equilib¬
rium measure PKi-y) coincides with the corresponding Chebyshev polynomial P 2 ^ +
rs/2 for all s G Nq.

Proof. For s = 0 we have the result by symmetry. Let s > 1. Each polynomial P
of degree less than 2® is a linear combination of polynomials of the type

(P2« 1 +

. (^2 + ^)"'

V 2 /

V 2 /

V 2

ORTHOGONAL POLYNOMIALS FOR THE WEAKLY EQUILIBRIUM CANTOR SETS 5

with m S {0,1}. By Lemma [2^ P 20 + Ts / 2 is orthogonal to all polynomials of
degree less than 2®, so it is Q 2 ‘- □

By (|2.1L we immediately have
Corollary 2.9. (52=+i = Qi-* ~ (1 ~ 2 7s_|_i)r^/4 for s G Nq.

3. Some products of orthogonal polynomials

So far we only obtain 2® degree orthogonal polynomials. We try to find Qn for
other degrees. By Corollary 12.91 since f Q 2 “+i'^Ma'( 7 ) = 0, we have

(3.1) 11^2^1"= J QldfiK(-y) = il-2js+i)r^s/^

and

(3.2) Q2e+i = QI'® ~ IIQ2''|Pj G Nq.

Our next goal is to evaluate f A for A-polynomial of the form

(3.3) A = (Q2.„ )*" )*”-! ... (Q2n ,

where s„ > s„_i > ... > si > 0 and fi, * 2 ,..., G {1, 2}.

The next lemma is basically a consequence of (|3.2I1 .

Lemma 3.1. Let A he a polynomial satisfying (13.31) . Then the following proposi¬
tions hold:

(^) 2 then J' A dg,i^ ||^ 2 ®®^ || f? 2 Sn—i * *' ^ 2^1 ^Mic( 7 ) •

(b) If in = in-1 = ... = in-k = 1 and in-k-i = 2 where fc G {0,1,..., n - 2},
then

j A dllx{'y) — ||Q2“" II J Q2‘‘n-k-2 ' ' ' Q2 ‘‘i dl-lK('y );
provided Sn = s„_i + 1 = Sn -2 + 2 = • • • = + fc + 1.

(c) Ifik = I and Sk > Sfc_i+2 for some fc G {2,3,..., n}, then J A d/rx( 7 ) = 0.

W // 7 = = . ■. = = 1 /»r .ome t C {1.2,... . tken j A = «■

Proof. (a) Using (1321), we have = Q 2 ^n+i + ||Q 2 '®n||^. The result easily
follows since the degree of ’ ’ ’ ^ 2=1 is i®ss than 2®"+^.

(b) Here A = Q 2 ^n Q 2 '®~-i ■ ■ ■ Q 2 ‘’n-kQl,^.k-i ■ P with P = ■'' Q 2 U ■

Observe that the degrees of the first fc + 2 polynomials are consecutive and
deg P < 2®"-'®-2+2 < 2 ®"-'®. We apply (13.21) repeatedly. First, Q^on-fc-i =
Q 2 Sn-k + ||(52'®"-'=-l|P. Then Q 2 Sn-k -*:-l ~ Q 2 ‘‘n-k + l + ||Q 2 '®»‘-'“|P T
IIQjsn-fe-iIP ■ Q 2 ‘n-k. After fc + 1 steps we write A in the form ((52®®"+! +
IIQ 2 '®" IP + >C) P, where £ is a linear combination of the polynomials Q 2 ^n ,
Q 2 ‘n (52'®n-i, • • • , Q 2 ‘ri-i ''' <52'®"-'® ■ By orthogonallty, all terms vanish
after integration, except ||Q 2 '>"|pP, which is the desired conclusion.

6

GOKALP ALPAN AND ALEXANDER GONCHAROV

(c) Let US take the maximal k with such property. Repeated application
of (a) and (6) enables us to reduce f A to C f Ai d^K{-y) with

C > 0 and Ai = Q 2 ‘-m ■ ■ ■ Q 2 ‘>k ■ R, where R = ■ ■ ■ Q 2 *i

deg i? < 2®'“-!+^ < 2®'=. Comparing the degrees gives the result.

(d) Similarly, f Ad^iK{-y) = C ■ f Q 2 ‘‘k ■ ■ ■ Q 2 '=idMAr( 7 ) = 0.

□

Theorem 3.2. For A—polynomial given in (EH) with So := —l,*r!,+i := 2, let
Ck = (4-1)®'“"®''-!"^ and c = nLiCfe- Then f Adfij^(^) = c-OLi

Proof. First we remark that c G {0,1} and c = 0 if and only if ii = 1 or ife = 1
for some k G {2, 3, • • • ,n} with Sk > Sk-i + 1, that is just in the cases (c) and (d)
above.

Secondly, the procedures (a) — (d) of Lemma [ST] allow to find / AdpLx{-)) for all
admissible values of iik)k=i {sk)k=i- Consider the string I = {i„, in-i, ■ ■ ■ , *i}.
If = 2 then we use the procedure (a). Suppose *„ = 1. If all other it = 1 then
/ A dyLx(j) = 0, by (d). Otherwise we have a string {I, • • • , I, 2, • • • }. Here we check
the values Sk- If they are consecutive (including the first appearance of 2), then we
use (&) and get the term ||( 52 »n in front of the new integral. Otherwise,

by (c), the integral is zero.

Thus, in the case of non zero integral, we decompose I into substrings of the
types {2}, {I, 2}, • • • , {I, • • • ,1,2} without gaps between consecutive Sk in the same
substring. If there are two or more consecutive substrings {2} then gaps between
corresponding Sk are allowed. Let k with A: < n be the index of the first term in
Ik = {1, • • • ,1, 2}. Then ik+i = 2 and, by (6), Ik brings in the term \\Q 2 k |P into
the common product. □

Corollary 3.3. For A—polynomial given in (13.311 . let A = Ai ■ , so Ai con¬

tains all terms of A except the last. Suppose = *2 = 2. Then f Adfj,x(j) =
||Q2n IP / dp,x{j).

We will represent Qn in terms of B-polynomials that are defined, for 2"* < n <
2 ?™+! m G No, as

Bn = iQ2r^y-'{Q2^-^y^-A..{Qiy\

where ik G {0,1} is the fc—th coefficient in the binary representation n = im 2™ +

• • • + iQ.

Thus, Bn is a monic polynomial of degree n. The polynomials B( 2 fe+i). 2 « and
i3(2j+i).2™ are orthogonal for all j, k,m, s G Nq with s y m. Indeed, if minjm, s} =
0 then f i3(2fc+i).2* B(^ 2 j-i-i)- 2 ^ = 0, since one polynomial is symmetric about

X = 1/2, whereas another is antisymmetric. Otherwise we use Lemma |3.II (d). By
(a), we have

m m

l|Sn|P = n 11^2^11'*''= n

k—0 k—0,ik^0

Theorem 3.4. For each n G N, let n = 2®(2fc + l), the polynomial Qn has a unique
representation as a linear combination of B 2 ‘,B^. 2 ‘ ..., B(^ 2 k-i). 2 ‘ ■, B(^ 2 k+i)- 2 ‘ ■

Proof. Consider P = aoB 2 ‘ + aiB^. 2 - + ... + ak-iB( 2 k-i). 2 ‘‘ + -B( 2 fc+i). 2 '’, where
are chosen such that P is orthogonal to all B{ 2 j+i) 2 ‘‘ with j = 0,1,..., k— 1.

ORTHOGONAL POLYNOMIALS FOR THE WEAKLY EQUILIBRIUM CANTOR SETS

7

This gives a system of k linear equations with k unknowns (aj)^Tg- The determi¬
nant of this system is the Gram determinant of linearly independent functions
(i3(2j+i)20j=o • Therefore it is positive and the system has a unique solution. In
addition, as was remarked above, P is orthogonal to all B(^ 2 j+i)- 2 "' with m ^ s.
Thus, P is a monic polynomial of degree n that is orthogonal to all polynomials of
degree < n, so P — Qn- □

Corollary 3.5. The polynomial Q 2 ‘{ 2 k+i) is a linear combination of products of
the type Q 2 ‘m-i ■ ■ ■ so the smallest degree of Q 2 ‘‘j in every product is 2®.

To illustrate the theorem, we consider, for given s £ Ng, the easiest cases with k <
2. Clearly, Q 2 ^ = 52 ® ■ Since P 3 . 2 ® = Q 2 ‘^Q 2 ^+^i we take Q 3 . 2 ® = o,oQ 2 ^ +Q 2 '>+iQ 2 ®j
where ag is such that f < 53 . 2 * Q 2 » dyiK{-y) = 0. By Lemma [5TT1

Q3-2‘ — (52=+iQ2«

\\Q2^+A?

II02HP

Q2‘ ■

Similarly, 55 . 2 = = Q 2 ‘Q 2 ^+‘^ and Q 3 . 2 ^ = aoQ 2 '» + aiQ 2 ^+iQ 2 ‘ + Q 2 ‘Q 2 ^+'^ with

IIQ2=+HP wqaa

IIQ2«II^-||Q2 ^+i|P’ 1102^+1 IP'

Using (EH), all coefficient can be expressed only in terms of (7fc)^i. As k gets
bigger, the complexity of calculations increases.

Remark 3.6. In general, the polynomial Qn is not Chebyshev. For example, Qg =
Qi{Q 2 + oo) with flg = — least for small 71 , the polynomial Qaix) =

(x — l/2)(x^ — X + 71/2 -I- Op) increases on the first basic interval Ii.i = [0,Zi_i].
Here, Zi,i is the first solution of P 2 = —ri, so = (1 — \/l — 4ji)/2. If Q 3 is the
Chebyshev polynomial then, by the alternance argument, Q 3 (^i,i) = Q 3 (l), but it
is not the case.

4. Jacobi parameters

Since fJ'K('y) is supported on the real line, the polynomials (Qr!,)5^o satisfy a
three-term recurrence relation

Qn+l{x^ — (x bn-\-QQn(yX^ Qn—l(^xf 71 £ f^O-

The recurrence starts from Q-i := 0 and Qq = 1. The Jacobi parameters {on,
define the matrix

/

Oi

0

0 .

Oi

&2

02

0 .

0

02

b2

03 .

1 :

/

where t^Kiy) is is the spectral measure for the unit vector Ji and the self-adjoint op¬
erator J on which is defined by this matrix. We are interested in the analysis

of asymptotic behavior of Since /ia'( 7 ) is symmetric about x = 1/2, all

are equal to 1/2. It is known (see e.g. [5D]) that an > 0, ||Qn|| = oi • • • a„, which,
in turn, is the reciprocal to the leading coefficient of the orthonormal polynomial
of degree n.

In the next lemmas we use the equality f QnQmQn+mdpLK{-y) = ||Qn-i-m|P,
which follows by orthogonality of Qn+m to all polynomials of smaller degree.

GOKALP ALPAN AND ALEXANDER GONCHAROV

Lemma 4.1. For all s G No and fc G N we have

Q2’>(2k+l) — Q2‘ ■

\\Q2‘{2k-l) IP

Q2‘*(2fc-1) ■

NO II ^

Proof. Consider the polynomial P = Q 2 = ■ Q 2 ^+^k — li^^^^^‘Vlp*32''(2fe-i)- It is a
monic polynomial of degree 2'*(2fc + 1). Let us show that P is orthogonal to Qn for
all n with 0 < n < 2®(2fc + 1). This will mean that P = Q 2 ‘*( 2 fc+i)-

Suppose 0 < n < 2®(2fc — 1). Then orthogonality follows by comparison of the
degrees.

If n = 2^{2k — 1) then f P Qn dfixi-y) = 0 due to the choice of coefficient of the
addend in P and the remark above.

Let 2®(2fc — 1) < n < 2®(2A: + 1). We show that f Q 2 ^ Q 2 ‘+^k Qn dfJ.K(j) = 0. We
write k in the form k = 2^(2/ +1) with some q,l G Nq. In turn, n = 2'"(2p+ 1) with
m s. By Corollary 13.51 Q 2 ‘+^k is a linear combination of products of Q 2 ‘’j with
minsj = s + I + g in every product. Similarly for Qn, but here the smallest degree
is 2"^. Therefore, Q 2 '' Q 2 ‘+^k Qn is a linear combination of yl—polynomials and for
each yl—polynomial the exponent of the smallest term is I. By Lemma Fd.lI d'). the
corresponding integral is zero. □

Lemma 4.2. For all s G Nq and fc G N we have

22 2,22 2 _ 11,^ ||2

®2®(2fc+l) ®2®(2fc+l)-l ■ ■ ■ ®2=+ife+l + ®2'*+ife ®2=+ife-l ’ ’ ‘ ®2®+ifc-2® + l “ II|| •

Proof. By Lemma 14.11 and the remark above,

(4.2)

Let us show that

IIQ2''(2fc+l)lP — J QI^ Q2‘+^k^FK{'Y) —

IIQ:

2'»+ifcl

\\Q

2‘>{2k-l)

12 ■

J Q 2 '’ Q2'’+^k^FK{~f) — IIQ 2 ®

^IIQ

2^+^A; I

If A: = 2"*, we have this immediately, by Lemma IXTI aL

Otherwise, 2®+^A: = 2™(2^ + 1) with I G N and m > s + 1. Then, by Corollary
3.51 <52®+!fc is a linear combination of products Q 2 ‘q ■ ■ ■ Q 2‘3 '' ■ <52™ with Sj > to
except the last term. From here, where ^ Uj Aj is a linear

combination of type polynomials with si > to for each Aj. Therefore,

IIQ2''+ifell ~ ^ ^ ^3 J -^jQ2^

On the other hand,

/ Q 2 ® Q2‘‘+3k^h‘K{'f) ~ ^ ^ / AjQ2m Q2'>

By Corollary 13.31 this is ||Q 2 ®|P||Q 2 '=+ifc
Therefore, (lO) can be written as

||Q2»(2fc+l)|P

= iiQ2^r-

IIQ:

2^ + ife I

IIQ2‘>+ifeP I|i52»(2fc-1)

which is the desired result, as a„ = ||Qn|| / ||Q„_i||.

12 ’

□

Theorem 4.3. The recurrence coefficients (a„)^^ can be calculated recursively by
using Lemma \4-.S\ and (EH).

ORTHOGONAL POLYNOMIALS FOR THE WEAKLY EQUILIBRIUM CANTOR SETS 9

Proof. We already have ai = ||Qi|| and 02 = ||( 52 || / ||Qi||- Suppose, by induction,
that all Oi are given up to i = n. If n + 1 = 2® > 2 then

(4.3)

^n +1 —

IIQ 2

I |Q2®-i 11 ■ 02*'“ 1 + 1 ■ 02*“1+2 ■ ■ ■ O2* —1

where the norms of polynomials can be found by (EH).

Otherwise, n + 1 = 2®(2fc + 1) for some s G No and k G N. By Lemma [4.21 we
have

(4.4)

2 _ 2 _ 11^2*11 ®2*+ife ■ ■ ■ ®2*+ifc-2* + l

®n+l — ®2*(2fe+l) — 2 „2

“2*(2fc+l)-l ■ ■ ■ “2* + ifc+l

provided s 0. If s = 0 then the denominator in the fraction above is absent. This
gives On+i, since the recurrence coefficients are positive. □

In order to illustrate the theorem, let us consider the cases of small s.

If s = 0 then n + I = 2A: + 1 and 02 j ._|_2 = of — o^^.. Next, for s = 1 and s = 2,

„2 IIQ 2 IP — O 4 ;. 04;._^ ^2 IIQ 4 IP — o|j, Og ^_2 Og^_3

®4fc+2 “ 2 ’ ®8fc+4 “ 2 „2 „2 ’

“4;s+l “8fe+3 “8fc+2 “ 8 fc+l

Thus, Oi
etc.

2

02

vl -272

71, a§ = o? - oi.

04

7172 Vl -273

0.3 \/ 1“272 ’

2 2 2
ai = Of - a|.

Remark 4.4. If 7 „ < 1/4 for I < n < s and 7 ^ = 1/4 for n > s then K^j) = Eg =
( 2 P 2 */ts + 1)~^[—1, !]• Here (P 2 " +Tn/2)))To Chebyshev polynomials for Eg,

as is easy to check. Therefore Theorems 12.81 and 14.31 are applicable for this case as
well. For further information about Jacobi parameters corresponding equilibrium
measures of polynomial inverse images, we refer the reader to the article [9].

Remark 4.5. Suppose 7 ^ = 1/4 for n < IV with 2® < IV < 2®+^. Then oi = l /\/8
and 02 = 03 = • • • = 02 *+i_i = 1/4. In particular, if 7 „ = 1/4 for all n then o„ = 1/4
for all n > 2, which corresponds to the case of the Chebyshev polynomials of first
kind on [ 0 , 1 ].

47 '^

Lemma 4.6. Suppose jg < 1/6 for all s. For fixed s G No, let c = (md

^ ~ i+^i-ic • following inequalities hold with fc G Nq."

(a) //n = 2 ®( 2 fc+l) then

i||Q2*|P < ||Q2*|P < 0^••oL2* + l < IIQ2*ir

(b) If n = 2^{2k2) then

a

2

n

®n-2* + l

- IIQ 2 *IP

^o IIQ 2 *+^IP
- IIQ 2 *IP ■

Proof. Note that, if 7 s+i increases from 0 to 1/6, then c increases from 0 to 1/4
and C increases from 1 to 2. By EH and the definition of r^, we get

I 1 Q 2 ''+i||^ _ 2 2 1 ~ 27s+2 _

- 7s+i rg ^ -

(4.5)

IIQ 2 *IP

(1 — 27^+2)c||Q2''|P < ||Q 2 *|P/ 4 .

10

GOKALP ALPAN AND ALEXANDER GONCHAROV

We proceed by induction. For a fixed s G No, let fc = 0. Then we have at once

1? = II<32* IP and a^s+i ■ ■ ■ =

\\Q2

Suppose (a), (&) are satisfied for k < m. We apply Lemma [4.21 with fc = to + 1 :

a2*(2m+3) ■ ■ ■ ^2‘{2m+2) + l + a2*(2m+2) ‘ ’ ’ ®2*(2m+2)-2* + l = ll<32* || ,
where for the addend we can use (6) for k = m. Therefore,

,II<32*+i|P / „2 „2 ^ 11^ ||2

||(32*P — ®2*(2m+3) ■ ■ ■ ®2*(2m+2) + l — II V2* II ,

11 ^ 2 * 11 '-^

which is (a) for fc = to + 1 , by g3]) .

Next, we claim that

(4-6) a2*(2m+4) ■ ■ ■ a2»(2m+2) + l — ll<32*+i||

for TO G No. If TO = 2/ + 1 then we use Lemma [4.21 with s + 1 instead of s :

“ 2 *+i( 2 fc+i) ■ ■ ■ ® 2 *+ 2 fe+i + positive term = ||<32»+i|P,

which implies (14.61) . if we take fc = / + 1, as 2(2fc + 1) = 2 to + 4, 4fc = 2 to + 2.
Suppose TO is even. Lemma 14.21 now gives

positive term + 02 ^+ 2 ^ • • • a 2 *+ 2 fc- 2 *+i+i = ll<32*+i IP,

where we take fc = to/ 2 + 1. Thus, (14.61) holds true in both cases.

Putting together (a) for fc = to + 1 and (Irel) we get (&) for fc = TO + 1 . □

Theorem 4.7. Let 7 ^ < 1/6 for all s. Then lim aj. 2 *+n = a„ for j G N and

s—^oo

n G No- Here, oq := 0. In particular, lim inf a„ = 0.

Proof. We first show that lim 072 * = 0 for all j G N. Let j = 2*(2fc + 1) where

S—>-oo

k,l G No. For s > 0, the Jacobi parameters admit the following inequality by
Lemma l4.6f al:

(4.7) a2s+i(2fe+i) ■ • ■ a2*+'+ife+i — ll<32*+'ll ■

If i < s + I where i G No, we have 2®+*(2fc + 1) — 2* = 2*(2'*+^“*(2fc + 1) — 1).
Since 2®+*“®(2/c + 1) — 1 is a positive odd number, by Lemma 14.fil ab we have the
inequalities

2ll<32HI ^ Q'2s+q2fc+l)-2* ■ • ■ ®2*+'(2fe+l)-2‘+i + l * = 0, . . . , S + Z — 1.

We multiply these s + I inequalities side by side:

2 ll<3l|| ■ ■ ■ I I<32 *+‘-i II ^ fl2*+‘(2fc+l)-l ‘ ' ®2*+* + ifc+l

and use (I4J1) :

2 _ 2 2-+^||Q2*+-IP

S-2* a2*+q2fc+i)- ||Q2*+<-PP||Q2 *+<-IP---IIQiIP'

By (14.51) . the fraction above is bounded by 2 Thus, lim aj. 2 ^ = 0.

s—>-oo

If n = 1 then aj.2*+i = of — a|.2s —>■ af, which is our claim.

ORTHOGONAL POLYNOMIALS FOR THE WEAKLY EQUILIBRIUM CANTOR SETS 11

Suppose, by induction, that lim aj. 2 ‘+n = On for n = 0,1,..., m and all j G N.

S^OO

Let TO + 1 = 2P{2q + 1) where p,q G Nq. If 9 = 0 then j • 2® + to + 1 = j • 2^~p + 1,
so we get the case with n = 1. Thus, we can suppose g G N. Then j • 2® + to + 1 =
2P^2®+^“P(2fc + 1) + 2g + 1) and, for large enough s, we can apply Lemma IT^

2 2 2 j_ 2 2 _ 11 ^ ||2

®y2'’+m+l ‘ ‘ ‘ ^j-2^+m-2P + l + ®j-2'’+m-2P ‘ ‘ ’ ®j-2‘*+m+l-2P+i “ IIL/ZpII •

Here all indices, except the first, are of the form j • 2® + n with n < to + 1.
Therefore, by induction hypothesis, a|. 2 s+n —>■ Un as s —>■ 00 and

( lim a".

s—>-oo

l)«^

■ • a

m-2P + l

I z

~r Clrn—'l

'■^771+1-2^+1 — II <52? I

On the other hand, if we apply Lemma |4.2l to the number to + 1, then we get the
same equality with instead of hms_>.oo aj. 2 <!+m+i- Since all Uk are positive,

we have the desired result. □

Remark 4.8. Since lim inf a„ = 0, by [5], the set K{‘^) with 7 ^ < 1/6 has zero
Lebesgue measure.

5. WiDOM FACTORS

A finite Borel measure /r supported on a non-polar compact set AT C C is said
to be regular in the Stahl-Totik sense if lim ||Qn||" = Cap{K) where Qn is the

n—voo

monic orthogonal polynomial of degree n corresponding to p. It is known (see,
e.g. [m [21]) that the equilibrium measure is regular in the Stahl-Totik sense.
While IIQrill" / Cap{K) has limit 1, the ratio Wn = ||Qn|| / {Cap{K)Y may have
various asymptotic behavior. We call Wn the Widom factor due to the paper [22] .
These values play an important role in spectral theory of orthogonal polynomials
on several intervals.

n

Let E = [a,/3] \ [J(ai,/3i) where a,/3 G M and the intervals {ai,(3i) are disjoint

i=l

subsets of [a,/3]. Let dp = f{t)dt on E and (an)))Ti be the Jacobi parameters
corresponding to p. Then [5]

(5.1)

J log f{t)dpE{t) > -00

lim sup

n—>-oo

Ql ■ ■ -Qn
Cap{E)^

> 0 .

For further generalizations and different aspects of this result, see [ 5 |[S|[Z|[I 1 [I 1 ].

We already know that Cap{K{j)) = exp (X^^i 2“^ logy^). In terms of
( 7 /c)fci ■we can rewrite IIQ 2 HI as

(5.2)

Therefore,

(5.3)

\/l — 27 s+i

exp ( 2®^2 '"log 7 fe j.

fe=i

W 2 . =

■\/l — 27 s+i

2exp(Er=.+i2®-Mog7fe)

> \/ 2 ,

since 7 s < 1/4. The limit values 7 s = 1/4 for all s give the Widom Factors for the
equilibrium measure on [ 0 , 1 ].

Clearly, (15.311 implies that lim sup Wn > 0. If 7 s < 1/6 for all s then

(5.4)

W2« > \/6.

12

GOKALP ALPAN AND ALEXANDER GONCHAROV

Let US show that, in this case, liminf Wn > 0 .

Theorem 5.1. Let (Wn)^=i be Widom factors for /ir-(7) where 7s < 1/6 for all s.
Then

(a) liminflL2s =liminfWn.

s—^oo n—^oo

(b) lim sup Wn = oo.

n—^oo

Proof. (a) We show that Wn > for 2^* < n < 2'*+^. Let n = 2® + 2*^ +
... + 2®”“ with s > Si > S 2 > ... > Sm > 0. Then we decompose oi • • • a„
into groups (oi • • • 02 *) • (a 2 »+i ■ ■ • a 2 <*+ 2 n ) • • • (a 2 »+...+ 2 '’m-i+i ■ ■ • a„). For
the first group we have oi • • • 02 ^ = ||Q 2 '>||- For the second group we use
Lemma ri.df ai with n = 2® + 2®i : a 2 <!+i • • • a 2 <i+ 2 '>i > ||Q2n||/\/2. Similar
estimation is valid for all other groups. Therefore,

_ Oi • • • a2<i 02^ + 1 ••• a2*+2*l a2“H ^2'’™-i+l ’ ‘

Cap{K{'y)^‘‘ Cap{K{'^Y‘^ Cap{K(pfY‘'^

> W2«W2n •••lF2.m(\/2)-’",

which exceeds W2s(-\/3)"*, by ( 15 . 41 ) . From here, min2<i<„<;2*+i = W^^

and the result follows.

(b) Applying the procedure above to lF2s_i and taking the limit gives the
desired result.

□

In order to illustrate the behavior of Widom factors, let us consider some exam¬
ples. Suppose 7s < 1/6 for all s.

Example 5.2. If 7s —>■ 0 then Wn —t 00. Indeed, W2‘ > exp(i log 7^)-

Example 5.3. There exists 7s ^ 0 with IF„ —t 00. Indeed, we can take 72/c =
1/6, 72 fc-i = 1/fc.

Example 5.4. If 7s > c > 0 for all s then liminf„_>.oo IF„ < l/ 2 c.

Example 5 . 5 . There exists 7 with inf 7s = 0 and liminf„_>.oo IF„ < 00. Here we
can take 7s = 1/6 for s ^ Sk and 7sj, = Ifk for a sparse sequence {sk)’^i. Then
(W2«fc)^i is bounded.

6. Towards the Szego class

The convergence of the integral on the left-hand side of ( 15 . 11 ) defines the Szego
class of spectral measures for the finite gap Jacobi matrices. The Widom condition
on the right-hand side is the main candidate to characterize the Szego class for the
general case, see [smiiii.

The equilibrium measure is the most natural measure in the theory of orthogonal
polynomials. In particular, for known examples, the values lim sup associated
with equilibrium measures are bounded below by positive numbers. So we make
the following conjecture:

Conjecture 6.1. If a compact set K G ^ is regular with respect to the Dirichlet
problem then the Widom condition Wn 0 is valid for the equilibrium measure
fJ-K-

ORTHOGONAL POLYNOMIALS FOR THE WEAKLY EQUILIBRIUM CANTOR SETS 13

Concerning the Szego condition, one can conjecture that, in the case of measures
with non-polar perfect support K, the left-hand side of dSI]) can be written as

( 6 . 1 ) I{fi) := J \og{dfi/dfiK)dfiK{t) > -oo.

Indeed, for the finite gap case, this coincides with the condition in (15.IL since the
integral J \og{dfiK/dt)dfiK{t) converges. By Jensen’s inequality (see also Section 4
in i), the value /(/r) is nonpositive and it attains its maximum 0 just in the case
/i = fiK a.e. with respect to fiK- On the other hand, there are strong objections to
(IQI) . based on the numerical evidence from [la, where, for the Cantor-Lebesgue
measure ficL on the classical Cantor set Kq^ the Jacobi parameters (a^) were
calculated for n < 200.000. The Widom factors for such values behave as a bounded
below (by a positive number) sequence. Therefore, if we wish to preserve the Widom
characterization of the Szego class, the integral must converge, but, since

fiCL and fiKo are mutually singular, it is not the case.

References

[1] Alpan, G., Goncharov, A.: Two measures on Gantor sets. J. Approx. Theory. 186 , 28-32
(2014)

[2] Barnsley M.F., Geronimo, J.S., Harrington, A.N.: Infinite-Dimensional Jacobi Matrices As¬
sociated with Julia Sets. Proc. Amer. Math. Soc. 88(4), 625—630 (1983)

[3] Barnsley, M.F., Geronimo, J.S., Harrington, A.N.: Almost periodic Jacobi matrices associ¬
ated with Julia sets for polynomials. Comm. Math. Phys. 99(3), 303-317 (1985)

[4] Bellissard, J., Bessis, D., Moussa, P.: Chaotic States of Almost Periodic Schrdinger Operators.
Phys. Rev. Lett. 49, 701-704 (1982)

[5] Christiansen, J.S.: Szego’s theorem on Parreau-Widom sets. Adv. Math. 229, 1180-1204

( 2012 )

[6] Christiansen, J.S., Simon, B., Zinchenko, M.: Finite Gap Jacobi Matrices, IL The Szego
Class. Constr. Approx. 33(3), 365-403 (2011)

[7] Christiansen, J.S., Simon, B., Zinchenko, M.: Finite gap Jacobi matrices: a review.. American
Mathematical Society. Proceedings of Symposia in Pure Mathematics, p. 87-103 (2013)

[8] Dombrowski, J.:Quasitriangular matrices. Proc. Amer. Math. Soc. 69, p. 95-96 (1978)

[9] Geronimo, J.S., Van Assche, W.:Orthogonal polynomials on several intervals via a polynomial
mapping. Trans. Amer. Math. Soc. 308 , 559—581 (1988)

[10] Goncharov, A.: Weakly equilibrium Cantor type sets. Potential Anal. 40, 143-161 (2014)

[11] Heilman, S.M., Owrutsky, P., Strichartz, R.: Orthogonal polynomials with respect to self¬
similar measures. Exp. Math. 20, 238-259 (2011)

[12] Kruger, H., Simon, B.: Cantor polynomials and some related classes of OPRL. J. Approx.
Theory (2014), http://dx.doi.Org/10.1016/j.jat.2014.04.003

[13] Makarov, N.G., Volberg, A.: On the harmonic measure of discontinuous fractals. LOMI
preprint, E-6-86, Steklov Mathematical Institute, Leningrad (1986)

[14] Peherstorfer, F., Yuditskii, P.: Asymptotic behavior of polynomials orthonormal on a homo¬
geneous set. J. Anal. Math. 89, 113-154 (2003)

[15] Ransford, T.:Potential Theory in the Complex Plane. Cambridge University Press, (1995)

[16] Saff, E.B., Totik, V.: Logarithmic potentials with external fields. Springer-Verlag, New York
(1997)

[17] Simon, B.: Almost periodic Schrodinger operators: A Review. Adv. Appl. Math. 3, 463-490
(1982)

[18] Simon, B.: Equilibrium measures and capacities in spectral theory. Inverse Probl. Imaging

I, 713-772 (2007)

[19] Sodin, M., Yuditskii, P.: Almost periodic Jacobi matrices with homogeneous spectrum,
infinite-dimensional Jacobi inversion, and Hardy spaces of character-automorphic functions.

J. Geom. Anal. 7 (1997), 387-435

[20] Totik, V.: Orthogonal polynomials. Surv. Approx. Theory 1 , 70-125, (2005)

14

GOKALP ALPAN AND ALEXANDER GONCHAROV

[21] Widom, H.: Polynomials associated with measures in the complex plane. J. Math. Mech. 16,
997-1013 (1967)

[22] Widom, H: Extremal polynomials associated with a system of curves in the complex plane.
Adv. Math. 3, 127-232 (1969)

Department of Mathematics, Bilkent University, 06800 Ankara, Turkey
E-mail address: gokalp@feii.bilkent.edu.tr

Department of Mathematics, Bilkent University, 06800 Ankara, Turkey
E-mail address: goncha@fen.bilkent.edu.tr